Main Effect Comparisons
Statistical Method: Pairwise comparisons of estimated marginal means with Tukey adjustment for multiple comparisons.
Group Comparisons (Control vs Case)
No significant group differences detected.
Childhood obesity is a pressing public health crisis and a primary driver of adult obesity. The gut microbiome is a critical mediator of metabolic health, influencing energy harvest, insulin sensitivity, and systemic inflammation. This makes it a key target for therapeutic interventions. While dietary carbohydrates are a major energy source, their quality—specifically their digestibility—plays a more crucial role than quantity alone in shaping metabolic outcomes.
This project, a collaboration between Lurie Children’s Hospital and Abbott Nutrition, investigates how the gut microbiota of adolescents with and without obesity respond to different types of carbohydrates. It builds on previous findings that slowly digestible carbohydrates (SDCs) can reverse obesity-related phenotypes in animal models. However, human responses are not uniform. There is significant interpersonal variation in how gut microbes metabolize carbohydrates and produce short-chain fatty acids (SCFAs), which are key signaling molecules.
This study aims to understand the variation in microbial metabolic responses to fast- and slow-digestible carbohydrates. The primary objectives are:
This analysis focuses on the SCFA production profiles, examining how they differ by obesity status (case vs. control), carbohydrate type, and time. By understanding these dynamics, we can move towards personalized dietary interventions tailored to an individual’s unique microbiome.
The fecal metabolome was analyzed using targeted metabolomics at the DFI Host-Microbe Metabolomics Facility (DFI-HMMF). All compounds were validated through retention time and fragmentation comparison to established standards.
Short-chain fatty acids (acetate, butyrate, propionate, 5-aminovalerate, and succinate) were quantified using Gas Chromatography-Mass Spectrometry (GC-MS) after derivatization with pentafluorobenzyl bromide (PFBBr), as described by Haak et al. (2018) with modifications.
Metabolites were extracted from 100 mg of fecal material using 80% methanol with internal standards. The extracts were then derivatized. Briefly, 100 µL of extract was mixed with borate buffer, PFBBr in acetonitrile, and n-hexane, then heated at 65°C for 1 hour. The resulting hexanes phase, containing the derivatized SCFAs, was analyzed by GC-MS.
Statistical Methods: Data analysis was conducted in R (v4.4.1). Technical replicates were averaged prior to statistical analysis. Linear mixed-effects models were fitted using the lme4 package with subject as a random effect to account for repeated measures. Post-hoc pairwise comparisons employed estimated marginal means (emmeans package) with Tukey adjustment for multiple comparisons. Significance was set at α = 0.05, with Benjamini-Hochberg adjustment applied where multiple testing occurred.
Model diagnostics were performed to ensure assumptions of normality and homoscedasticity were met. Effect sizes are reported alongside p-values to provide clinical interpretability.
We begin by exploring the overall trends in the data to understand the main patterns of SCFA production across groups, carbohydrate types, and time.
Statistical Method: Descriptive statistics calculated using the rstatix::get_summary_stats() function, providing common measures of central tendency and dispersion for SCFA concentrations stratified by experimental conditions.
Statistical Method: Box plots with individual data points overlaid. Statistical comparisons performed using Wilcoxon rank-sum tests between groups, with significance denoted by symbols (ns: p > 0.05, : p ≤ 0.05, : p ≤ 0.01, : p ≤ 0.001). These are pairwise comparisons and were not adjusted for multiple testing.
Statistical Method: Line plots showing group means ± standard error of the mean (SEM) across time points, stratified by carbohydrate type. Error bars represent standard error of the mean.
To specifically quantify the change over the 48-hour incubation period, we calculated the delta (change) in SCFA concentration for each subject under each condition. This “response magnitude” analysis isolates the effect of the intervention over time.
Statistical Method: Box plots showing delta change distributions (48h - 0h) with Wilcoxon rank-sum tests comparing groups within each carbohydrate type. Individual data points represent subject-level responses. These are pairwise comparisons and were not adjusted for multiple testing.
Statistical Method: Two-way factorial ANOVA testing main effects of group and carbohydrate type, plus their interaction, on delta change values. P-values adjusted using Benjamini-Hochberg method to control false discovery rate across multiple analytes.
The analysis of the response magnitude (delta) shows that while carbohydrate type significantly influences the degree of change, this effect is consistent across both Case and Control groups, as indicated by the non-significant interaction term for most analytes.
Given the minimal group differences observed, we examine the carbohydrate effects by combining Control and Case groups to increase statistical power and focus on the primary experimental manipulation.
Statistical Method: Box plots showing combined group data with Wilcoxon rank-sum tests comparing each carbohydrate type against the no-carbohydrate control. P-values adjusted using Benjamini-Hochberg method to control false discovery rate.
Interpretation: By combining groups, we achieve greater statistical power to detect carbohydrate-specific effects. This analysis reveals the primary metabolic impact of different carbohydrate sources on microbial SCFA production, independent of obesity status.
Ratios between SCFAs can provide deeper insight into the metabolic balance of the gut microbiome. For example, the acetate-to-butyrate ratio can reflect the balance between inflammatory and anti-inflammatory potential. Here, we calculate key ratios and analyze them.
Statistical Method: Box plots comparing SCFA ratio distributions between groups and carbohydrate types. Ratios calculated as simple quotients with infinite values converted to missing data for statistical analysis.
To account for the repeated measures design (measurements at 0h and 48h from the same subjects), we use linear mixed-effects models with subject as a random effect. This is the most appropriate statistical approach for this data.
The mixed-effects models confirm that carbohydrate type is a major driver of the metabolic balance (ratios), and significant interactions with time and group are also observed for certain ratios.
Here we present the comprehensive statistical analysis using linear mixed-effects models for the absolute SCFA concentrations. This approach correctly models the data structure with subject as random effect, providing the most reliable results. Model diagnostics were performed to ensure assumptions were met.
The following sections present detailed results for each SCFA, highlighting the key findings from the mixed-effects models:
The results show highly significant effects for carbohydrate_type, timepoint_hr, and their interaction, confirming that both factors are strong drivers of SCFA production. The comprehensive mixed-effects modeling reveals the complex interactions between group, carbohydrate type, and time in determining SCFA responses.
Statistical Method: Following significant omnibus effects in the mixed-effects models, we conducted post-hoc pairwise comparisons using estimated marginal means (EMMs) with the emmeans package. This approach provides model-based comparisons that maintain the original error structure and account for random effects, with Tukey adjustment for multiple comparisons to control family-wise error rate.
Statistical Method: Pairwise comparisons of estimated marginal means with Tukey adjustment for multiple comparisons.
No significant group differences detected.
Statistical Method: Simple effects analysis to decompose significant interactions. Group differences are tested within each carbohydrate type, and carbohydrate effects are tested within each group, using Tukey-adjusted pairwise comparisons.
No significant group differences within specific carbohydrate types.
The emmeans post-hoc analysis provides precise effect size estimates with confidence intervals, enabling clinical interpretation of the magnitude of SCFA differences. Significant contrasts indicate biologically meaningful differences in microbial fermentation capacity between experimental conditions.
A key goal of this project is to assess interpersonal variation. The following visualizations highlight the differences in SCFA production trajectories from one subject to another, revealing the high degree of personalization in microbiome metabolism.
Statistical Method: Line plots showing individual subject trajectories (thin lines) overlaid with group means (thick lines). This visualization reveals interpersonal heterogeneity in SCFA production responses, with no formal statistical testing applied.
Statistical Method: Heatmap visualization with square-root transformation of concentration values to improve color scale discrimination. Each row represents an individual subject, ordered by experimental group, with color intensity indicating SCFA concentration magnitude.
This analysis provides a comprehensive overview of SCFA production in response to different carbohydrates in adolescents with and without obesity. The reorganization of the analysis into thematic sections clarifies the main findings and their implications.
Key Findings:
Dominant Effect of Carbohydrate Type and Time: The primary drivers of SCFA production in this ex vivo model were the type of carbohydrate supplied and the incubation time. Both slow and rapid digestible carbohydrates led to significant increases in all measured SCFAs over 48 hours compared to the no-carbohydrate control. This was confirmed by both the temporal analysis and the robust statistical significance in the mixed-effects models.
Subtle Differences Between Case and Control Groups: While there were some minor differences in baseline SCFA levels, the overall response to the carbohydrate interventions was remarkably similar between the case (obesity) and control groups. The mixed-effects models showed no significant main effect of group or major interactions involving group, suggesting that the fundamental capacity of the gut microbiota to ferment these carbohydrates is not dramatically altered in the context of obesity in this cohort.
SCFA Ratios Reveal Deeper Metabolic Shifts: The newly added SCFA ratio analysis, including the mixed-effects models, provided a more nuanced view. Carbohydrate type significantly altered the metabolic balance, such as the acetate-to-butyrate ratio. These shifts are critical, as they may have different downstream effects on host health (e.g., inflammation, energy regulation) even if the absolute concentration changes are similar. The slow-digestible carbohydrate tended to produce more favorable ratios (e.g., lower acetate:butyrate).
Pronounced Interpersonal Variation: A central finding, highlighted in the reorganized structure, is the high degree of subject-to-subject variability. The individual trajectory and heatmap visualizations clearly demonstrate that some individuals are consistently high or low SCFA producers, and their response patterns to different carbohydrates are unique. This underscores the need for personalized approaches in nutrition, as a “one-size-fits-all” dietary intervention is unlikely to be effective.
Limitations:
Revisions and Improvements:
The type of carbohydrate available for fermentation is a powerful modulator of gut microbial SCFA production, influencing both the absolute amount and the metabolic balance of these key molecules. While the overall fermentation capacity appears similar between adolescents with and without obesity in this study, the pronounced interpersonal variation in metabolic responses is a critical finding. This highlights the immense potential and necessity of developing personalized, microbiome-targeted nutritional strategies for managing obesity and improving metabolic health. Future work should focus on identifying the specific microbial taxa and functional genes that drive these individualized responses.
## R version 4.4.1 (2024-06-14)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 22.04.4 LTS
##
## Matrix products: default
## BLAS: /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
## LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.20.so; LAPACK version 3.10.0
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8 LC_MONETARY=en_US.UTF-8
## [6] LC_MESSAGES=en_US.UTF-8 LC_PAPER=en_US.UTF-8 LC_NAME=C LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## time zone: America/Chicago
## tzcode source: system (glibc)
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] DT_0.33 kableExtra_1.4.0 emmeans_1.11.2 lmerTest_3.1-3 lme4_1.1-37 Matrix_1.7-1 RColorBrewer_1.1-3 viridis_0.6.5
## [9] viridisLite_0.4.2 rstatix_0.7.2 ggpubr_0.6.1 data.table_1.17.8 janitor_2.2.1 readxl_1.4.5 lubridate_1.9.4 forcats_1.0.0
## [17] stringr_1.5.1 dplyr_1.1.4 purrr_1.1.0 readr_2.1.5 tidyr_1.3.1 tibble_3.3.0 ggplot2_3.5.2 tidyverse_2.0.0
##
## loaded via a namespace (and not attached):
## [1] tidyselect_1.2.1 farver_2.1.2 fastmap_1.2.0 TH.data_1.1-2 digest_0.6.37 timechange_0.3.0 estimability_1.5.1
## [8] lifecycle_1.0.4 survival_3.8-3 magrittr_2.0.3 compiler_4.4.1 rlang_1.1.6 sass_0.4.10 tools_4.4.1
## [15] utf8_1.2.6 yaml_2.3.10 knitr_1.50 ggsignif_0.6.4 labeling_0.4.3 htmlwidgets_1.6.4 xml2_1.3.8
## [22] abind_1.4-8 multcomp_1.4-26 withr_3.0.2 numDeriv_2016.8-1.1 grid_4.4.1 xtable_1.8-4 scales_1.4.0
## [29] MASS_7.3-63 dichromat_2.0-0.1 cli_3.6.5 mvtnorm_1.3-2 rmarkdown_2.29 ragg_1.4.0 reformulas_0.4.0
## [36] generics_0.1.4 rstudioapi_0.17.1 tzdb_0.4.0 minqa_1.2.8 cachem_1.1.0 splines_4.4.1 parallel_4.4.1
## [43] cellranger_1.1.0 vctrs_0.6.5 boot_1.3-31 sandwich_3.1-1 jsonlite_2.0.0 carData_3.0-5 car_3.1-3
## [50] hms_1.1.3 pbkrtest_0.5.3 Formula_1.2-5 crosstalk_1.2.1 systemfonts_1.2.3 jquerylib_0.1.4 glue_1.8.0
## [57] nloptr_2.1.1 codetools_0.2-20 stringi_1.8.7 gtable_0.3.6 pillar_1.11.0 htmltools_0.5.8.1 R6_2.6.1
## [64] textshaping_1.0.1 Rdpack_2.6.2 evaluate_1.0.3 lattice_0.22-6 rbibutils_2.3 backports_1.5.0 broom_1.0.8
## [71] snakecase_0.11.1 bslib_0.9.0 Rcpp_1.1.0 svglite_2.1.3 coda_0.19-4.1 gridExtra_2.3 nlme_3.1-166
## [78] xfun_0.52 zoo_1.8-12 pkgconfig_2.0.3